PROJECT SUMMARY/ABSTRACT Autism Spectrum Disorders (ASD) are complex disorders manifested by qualitatively atypical social communication skills and an aberrant behavioral repertoire that vary in severity across individuals. We lack neurobiologically-grounded predictors of autism in the general population. Our studies seek to fill this critical gap in our knowledge about neurobiologically-grounded quantitative signatures that precede manifestations of ASD in toddlers recruited from the general population. We aim to (i) apply advanced computational analytic techniques to formally chart the emergence of atypical developmental trajectories, and (ii) uncover and validate neurobiologically-grounded, clinically meaningful subtypes predictive of future risk for atypical development, revolutionizing brain imaging in young children. In our previous work we have discovered that head movements during functional MRI provide an abundant source of useful movement data whose statistical features are linked to clinical and cognitive outcomes in children and adults diagnosed with ASD. Our recent studies have revealed that quantitative signatures of atypical learning trajectories can be detected as early as 1-2 months in infants at high familial risk for developing ASD. Atypical functioning of the sensorimotor system has deleterious functional consequences across diverse domains of learning and development and may contribute to ASD manifestations, in toddlers screened prospectively in the general population. Using data from the NIH-funded National Database for Autism Research (NDAR) we will test whether atypical movement variability during MRI scans during the 2nd year of life in N=212 toddlers from the general population is predictive of ASD or non-ASD outcomes (vs. typical development, TD) ascertained during the 3rd year. We will rigorously quantify key kinematic parameters during MRI scans acquired in toddlers ages 12-24 months according to different conditions, including sleeping or resting, while language is presented to sleeping toddlers, and also during a socially-orienting scan. We hypothesize that deleterious, context-incongruent signatures during the 2nd year of life in toddlers will be related subsequently to greater ASD manifestations at 36-48 months. Machine learning algorithms will be used to classify ASD, non-ASD, and TD toddlers. The overall goal of these studies is to illuminate the neurobiological basis of sensorimotor variability in toddlers from the general population and to establish that sensorimotor signatures are part and parcel of the child?s future ASD diagnosis, a finding which will have profound, transformative implications for neuroimaging methods in young children. This knowledge will provide new, early mechanistic insights into the basis of such associations recently established in children, adolescents, and adults with and without ASD, as well as in human infants, and advance Research Priorities of the NIMH.